spacemut
Spatial analysis of germline mutation patterns.
This page containts data and code from our manuscript. We developed a computational approach that employs co-variation of mutation frequencies along the genome to extract mutational processes operating in the human germline. Toward that goal, we implementated volume-regularized NMF (vrnmf), a powerful technique to reconstruct non-negative sources from their observed mixtures. Analysis of large-scale whole genome sequencing TOPMed dataset reveals 4 strand-independent and 5 strand-dependent processes. For 7 of which we provided biological associations or interpretations.
Data
Below we provide description of data files for TOPMed and gnomAD datset.
TOPMed dataset
TOPMed_10kb.txt contains estimated mutation frequency of 192 mutation type across 263,870 non-intersecting 10 kb genomic window. Each row of a matrix contains information on a window: chromosome, start and end positions and 192 mutation frequencies.
Standard deviation normalization
Before inference of mutational components each mutation rate was normalized to its standard deviation (sd) across genomic windows. Below are presented tracks of mutational components for sd-normalized mutation rates (these tracks are used in the paper):
TOPMed_10kb_spectra_sd.txt contains spectra of mutational components. Each column contains a vector of relative probabilities to generate 192 sd-normalized mutation types for a mutational component.
TOPMed_10kb_intensities_sd.txt contains intensities of mutational components. Each row represents a genomic window and contains information on its chromosome, start and end positions, as well as window-average intensities of 14 sd-normalized mutational components.
Original scales
Below are spectra and intensities of mutational components converted in the original scales of mutation rates (by multiplying to mutation type-specific standard deviation):
TOPMed_10kb_spectra.txt contains spectra of mutational components. Each column contains a vector of relative probabilities to generate 192 mutation types for a mutational component.
TOPMed_10kb_intensities.txt contains intensities of mutational components. Each row represents a genomic window and contains information on its chromosome, start and end positions, as well as window-average intensities of 14 mutational components.
gnomAD dataset
gnomAD_100kb.txt contains estimated mutation frequency of 192 mutation type across 26,625 non-intersecting 100 kb genomic window.
Standard deviation normalization
gnomAD_100kb_spectra_sd.txt contains spectra of 12 sd-normalized mutational components.
gnomAD_100kb_intensities_sd.txt contains intensities of 12 sd-normalized mutational components.
Original scales
gnomAD_100kb_spectra.txt contains spectra of 12 mutational components.
gnomAD_100kb_intensities.txt contains intensities of 12 mutational components.
(gnomAD dataset reveals only 12 components compared to 14 components of TOPMed dataset due to more than order of magnitude smaller sample size)
Vrnmf inference of mutational processes from TOPMed dataset
Inference of mutational components from TOPMed dataset can be reproduced with a few simple steps using vrnmf R package.
Following installation of vrnmf and spacemut R packages (below), one can proceed to a guide through application of vrnmf to inference of mutational processes in germline.
Installation instructions
install.packages("devtools")
devtools::install_github("kharchenkolab/vrnmf")
library(vrnmf)
A detailed guide through vrnmf can be found here.
We also recommend to install spacemut R package that provides a set of routines to visualize and interpret inferred mutational components.
install.packages("devtools")
devtools::install_github("hms-dbmi/spacemut")
library(spacemut)
Available code
Code of simulations of mutational processes is available here: http://pklab.med.harvard.edu/ruslan/spacemut/simulations_topmed.R
Other code is available upon the request.